1. No category

CLARIN-PL
CLARIN-PL – a Polish Language
Technology Infrastructure for the Users
Maciej Piasecki
Wrocław University of Technology
G4.19 Research Group
maciej.piasecki@pwr.wroc.pl
2014-11-22
Users make problems
PALC 2014
Łódź
2014-11-22
CLARIN-PL
Users make all software systems imperfect.
However if a software system is not used, it
does not exist.
Who can use language technology?
Basic Notions
PALC 2014
Łódź
2014-11-22
CLARIN-PL
 Language Technology (LT)




language resources and tools
robust in terms of quality and coverage
multipurpose
component based
 Language Technology Infrastructure
 a software framework (architecture or platform)
 for combining language tools with language resources into
processing chains (or pipelines)
 the defined processing chains are next applied to language
data sources
 interoperability, also with the external systems
LT in Humanities and Social
Sciences: Barriers
PALC 2014
Łódź
2014-11-22
CLARIN-PL
 Physical – language tools and resources are not
accessible in Internet
 Informational – descriptions are not available or there is no
means for searching
 Technological – lack of commonly accepted standards for
LT, lack of a common platform, varieties of technological
solutions, insufficient users’ computers
 Related to knowledge – the use of LT requires
programming skills or knowledge from the area of natural
language engineering
 Legal – licences for language resources and tools (LRTs)
limit their applications
LTI for H&SS: Lowering Barriers
PALC 2014
Łódź
2014-11-22
CLARIN-PL
 CLARIN
 ERIC consortium of several countries
 member countries contribute parts of the LTI
 CLARIN Mission
 Lowering the barriers for LT in Humanities & Social Sciences
(H&SS)
 integration of different LT components into one interoperable
system
 one sign on and one login into the distributed infrastructure
 common standards
 common licences and promotion of the open access
 installation-free, web-based user interface
Different ways to LTI
PALC 2014
Łódź
2014-11-22
CLARIN-PL
 Bottom-up
 a collected offer approach
 based on linking together the already existing Language
Resources and Tools
 focused on accessibility, technical interoperability and
processing chains
 Top-down
 based on user-centred design paradigm
 research applications for H&SS are a starting point
 Bi-directional
 linking of Language Resources and Tools
 combined with the development of research applications
Bi-directional LTI development
PALC 2014
Łódź
2014-11-22
CLARIN-PL
 Idea
 development of the necessary elements
 a distributed network infrastructure
 basic LT processing chain
 combined with user-centred approach based on the
development of research applications
 Characteristic features
 a metaphor of the Agile-like light weight software designing
method
 close co-operation with key users from the H&SS domain
 application development stimulates the construction of
technical fundaments
 inspirations and identification of the further user needs
CLARIN-PL
PALC 2014
Łódź
2014-11-22
CLARIN-PL
 Polish scientific consortium
 Wrocław University of Technology, G4.19 Research Group
 Institute of Computer Science, Polish Academy of Science
 Polish-Japanese Institute of Information Technology, Chair of
Multimedia
 University of Łódź, PELCRA group at Chair of English Language
and Applied Linguistics
 Institute of Slavic Studies, Polish Academy of Science
 Wrocław University
 Goal: implementation of the Polish part of the CLARIN
ERIC LTI
 Generously financed by the Polish Ministry of Science and
Higher Education (about 4 millions Euro for three years)
 An example of the bi-directional approach
CLARIN-PL structure
PALC 2014
Łódź
2014-11-22
CLARIN-PL
 Context
 many basic LRTs for Polish were still lacking at the start of
CLARIN-PL
 Deeper technological barrier
e.g. the lack of a robust dependency parser for Polish
 Pillars:
 CLARIN-PL Language Technology Centre
www.clarin-pl.eu
the Polish node of the CLARIN distributed infrastructure
 Complete set of basic LRTs for Polish
 Research applications for H&SS – first created for key users
and selected H&SS sub-domains.
CLARIN-PL Langauge
Technology Centre: bottom-up
PALC 2014
Łódź
2014-11-22
CLARIN-PL
 B-type centre, located in Wrocław University of Technology
 based on modified D-Space system from Lindat (Czech CLARIN)
 Distributed authorisation
 linked to the national identity federation
 one sign-on, one login
 Proper repository system supporting
 persistent identifiers for resources and tools,
 CMDI meta-data format
 Interface for Federated Content Search
 On meta-data and content of corpora
 Depositing service for researchers from H&SS
 focused on LRTs
 adherence to all CLARIN specifications about standards and protocols
 Web Services for LRTs:
 the basic processing chain of Polish
 Flexible composition of the specialised processing chains
 SOAP & REST interfaces
 An active K-type centre in several areas
CLARIN-PL: Bottom-up
http://nlp.pwr.edu.pl/synat
PALC 2014
Łódź
2014-11-22
CLARIN-PL
Bi-directional: bottom-up part
PALC 2014
Łódź
2014-11-22
CLARIN-PL
 LRTs and LRT chains can be useful …
 if the required tools and resources exist,
 and, they are robust!
 What is the minimal set of LRTs?
 What kind of LRTs can be called robust?
 automated applications in H&SS seem to require high quality
of language tools and mostly large coverage of resource
 BLARK – The Basic Language Resource Kit
 “the minimal set of language resources that is necessary to do
any precompetitive research and education at all” (Krauwer,
2003) and also basic processing chains
 possible reference point to compare LRTs for different
languages
CLARIN-PL: language resources
PALC 2014
Łódź
2014-11-22
CLARIN-PL
 Good starting point, e.g.
 a huge National Corpus of Polish (1 billion tokens)
 plWordNet 2.0 – a very large wordnet for Polish
 Korpus Politechniki Wrocławskiej – an open Polish corpus
with rich annotation
 Main goals
 completing the construction of selected resources
 building bi-lingual resources and specialised corpora
facilitating the envisaged needs of H&SS
 Bilingual resources crucial for interoperability
 Large number of language pairs vs limited funds
 Priority given to Polish-English resources
CLARIN-PL: selected resources
in development
PALC 2014
Łódź
2014-11-22
CLARIN-PL
 plWordNet 3.0
 a comprehensive description of the Polish lexico-semantic system
(~200 000 lemmas, ~280 000 senses)
 mapping to enWordNet – an expanded Princeton WordNet 3.1
 A large lexicon of the Multi-word Expressions
 described with the minimal constraints on their lexico-syntactic
structures
 linked to plWordNet
 NELexicon 2.0 - ~2.5 million distinct PNs, semantically classified
 Dynamic lexicons – tools for automated expansion of the manual core
 A large semantic valency lexicon for Polish predicative lexical units
 Corpora:
 a transcribed training-testing Polish speech corpus, conversational corpus
 parallel corpora, historical Polish corpus of text news…
 Several systems for searching text and speech corpora
CLARIN-PL: language tools
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
PALC 2014
Łódź
2014-11-22
CLARIN-PL
Segmentation into tokens and sentences
Morphological analysis
Morphological guessing of unknown words (both without context and context
sensitive)
Morpho-syntactic tagging
Word Sense Disambiguation
Chunker and shallow syntactic parser
Named Entity Recognition and disambiguation
Co-reference and anaphora resolution
Temporal expression recognition
Semantic relation recognition
Event recognition
Shallow semantic parser
Deep syntactic parser with disambiguated output: dependency and constituent
Deep semantic parser
PALC 2014
Łódź
2014-11-22
CLARIN-PL: language tools
CLARIN-PL
 A generic set of morpho-syntactic tools for Polish that can
be adapted to a domain specified by the user
 Tools for the extraction of the semantic-pragmatic
information from documents and collections of documents,
e.g.
 keywords,
 semantic relations between text fragments
 and text summaries
 Web Services will be provided for all LRTs and systems
 already implemented: segmentation, morphological analysis,
tagging, chunking, Named Entity Recognition, and WSD
 accessible via REST or SOAP and described by CMDI
Web Service test
for Named Entity Recognition
PALC 2014
Łódź
2014-11-22
CLARIN-PL
Bi-directional - top-down part:
selection of applications
PALC 2014
Łódź
2014-11-22
CLARIN-PL
 Criteria




to cover a maximal variety of research areas
but also to co-operate first with the most active users
matching the available LT for Polish
a few application but broadening our understanding of the
domain
 First applications
 Spokes – a search system for the corpus of conversational
data (users from inside of CLARIN-PL)
 A system for collecting Polish text corpora from the Web
 A open textometric and stylometric system focused on Polish
 Semantic text classification for sociology
 Literary Map
System for collecting Polish text
corpora from the Web
PALC 2014
Łódź
2014-11-22
CLARIN-PL
 Requests from users revealed gaps in the available
technology
 Existing corpus building systems were too sensitive to text
encoding errors found in the web
 A system for collecting Polish text corpora from the Web
had to be constructed:
 based on solutions developed in Masaryk University in Brno
 applies morphological analysis to detect texts including larger
number of errors
 Supports semi-automated extraction of texts from blogs
Open textometric and stylometric
system
PALC 2014
Łódź
2014-11-22
CLARIN-PL
 Several textometric and stylometric tools available
 But not designed for languages of rich inflection like Polish
 Enabling the use of features defined on any level of the
linguistic structure:
 from the level of word forms
 up to the level of the semantic-pragmatic structures.
 Re-use of several existing components, e.g. Stylo
 Available as Web Application and a Web Service
 Stylometric techniques appear to be applicable in many
tasks of H&SS
 sociology (characteristic features that are for different
subgroups), political studies (similarity and differences
between political parties), literary studies …
Semantic text classification for
sociology
 Users: Collegium Civitas, Warsaw
 Initially:
 Text document classification according to the manually
annotated examples
 Finally:
 Whole system from corpus gathering to tuning machine
learning methods for the semantic classification of text
snippets
PALC 2014
Łódź
2014-11-22
CLARIN-PL
Semantic text classification for
sociology
1. Corpus building
2. Pre-processing


Text segmentation utilising the original structure
Morpho-syntactic tagging, parsing
3. Automated sample selection


Collection distribution
Clustering – different techniques
4. Manual annotation


Abstract definitions of semantic classes
Availability of open annotation editors
5. Training classifiers
6. Analysis of the results

Error estimation
PALC 2014
Łódź
2014-11-22
CLARIN-PL
GeTClasS – Generalised Text
Classification for Sociology
PALC 2014
Łódź
2014-11-22
CLARIN-PL
Literary Map
PALC 2014
Łódź
2014-11-22
CLARIN-PL
 Users
Digital Humanities Centre of The Institute of Literary Research
PAS)
 Idea
to identify all geographical names in the literary text (or a
corpus) and map them onto the geographical map
 Technical requirements
 Named Entity Recognition combined with geo-location
 PNs recognised in text must be grouped into expression
recognised by Google
 Recognition of semantic relations between non-spational PNs
and locations
 Parallel research on the method and its applications
Literary Map
PALC 2014
Łódź
2014-11-22
CLARIN-PL
PALC 2014
Łódź
2014-11-22
Conclusions
CLARIN-PL
 Application of LT to the research in Humanities & Social
Sciences seem to be much more challenging than in
commercial systems!
 LT for Polish achieved a stage in which valuable support
can be provided for research applications
 Bi-directional approach combines
 development of the basic, universal set of language tools and
resources
 with inspirations from the research applications
 Error monitoring and management in LT-based applications
is required
CLARIN-PL
Thank you very much for your attention!
www.clarin-pl.eu
Supported by the Polish Ministry of Science and Higher
Education [CLARIN-PL]
and the EU’s 7FP under grant agreement no 316097 [ENGINE]